Lung Cancer Detection from X-Ray Image Using Statistical Features*

Transcription

1 International Journal of Computing Academic Research (IJCAR) ISSN , Volume 4, Number 6 (December 2015), pp MEACSE Publications Lung Cancer Detection from X-Ray Image Using Statistical Features* Vinod Kumar 1, Ashu Gupta 2, Rattan Rana 3 and Kanwal Garg 4 1 Punjab Technical University, Kapurthala, India 2 Apeejay Institute of Management Technical Campus, Jalandhar, India 3 Kalpi Institute of Technology, Kalpi, Ambala, India 4 Kurukshetra University, Kurukshetra, India Abstract Lung cancer is a leading cause of mortality than any other type cancer in the world. Cancer in itself has some scary implications fall victimize to it, and 1.6 million pass away as a result of it. In India, the number of new cases improved from approximate 65,000 in 2009 to 90,000 in 2013, registering 15-20% increase annually. It is troublesome in India where people hesitate to consult doctors at the earliest or do not have access to them. In the modern era, the image processing mechanisms are used in several medical professions for advancing and enhancing the detection of lung cancer. Medical professionals look time as one of the important factor to discover the cancer in the patient at the earlier stage; which is very important for successful treatment. The current paper presents a methodology for exact diagnosing using statistical features from x-ray imaging. It is very difficult to detect quickly through x-ray as it contain noise. Thus, improve the x-ray images by using a filter as well as segmentation, threshold and edge detection approaches. Using first order histogram, the statistical features has been calculated. Keywords: Lung Cancer, Segmentation, Edge Detection, Statistical Features, Image Processing. Introduction Lungs are a unit set on the lateral sides of the body cavity and separated from one another by the bodily cavity because the left respiratory organ is physically smaller than the correction respiratory organ this can be as a result of the correction and left lungs exhibit some obvious structural variations as the center come keen on the left aspect of the body part cavity, the accurate respiratory organ is partitioned as the superior, middle, and inferior lobes by 2 fissures. The left respiratory organ combined with a medial surface indentation, submitted to as the internal organ impression that's approached by the center [1]. Lung illness refers to disorders that have an effect on the lungs. The respiratory area caused by respiratory organ illness that could forestall the body from obtaining enough gas [2]. Lung cancer may be an extremely aggressive and often fatal malignancy that originates within the epithelial tissue of the respiratory system. Smoking originates all respiratory organ cancers. Metastasis illustrate unfold of cancerous cells to alternative tissues, happens early within the course of the illness, creating a surgical cure unlikely for many patients [3]. X- ray * The paper was first presented in International Conference on Recent Trends in Computing and Information Technology organized by B.S Anangpuria Institute of Technology and Management, Faridabad, India. 178

2 imaging is that the quickest, most typical, and least costly diagnostic. Production of digital X-rays from pictorial radiographs is turning into a typical follow to maximize info and cut back the amount of rejected X- rays [4]. X-rays are a unit among the oldest sources of magnetic radiation used for imaging. The most effective use of X-rays in medical line is diagnosing the cancerous illness very beginning state. [2]. Methodology Analysis of tissue may be a method that is extremely self-made within the identification of medical pictures of abnormal respiratory organ, tissue and for this reason; the author describe concerning the x-rays and digital image process, which is that the means of recent identification defect as well as the illness. The steps to discover the cancer are:- (a) Cancelling the rips that seem white on the respiratory organs this could create mastics in detection the cancer by subtracting the background of the lung image from the cancer. (b) Dividing the respiratory organ that carries the cancer into 2 elements (normal and abnormal). Examining these with one another by histogram and applied math notations. (c) After specifying the partially hold the cancer in the image is processed by the subsequent steps:- A. Image Segmentation Image segmentation is that the most troublesome task. It is the process in image processing to grouping of image parts that exhibit similar characteristics, i.e. subdividing a picture into its constituent regions or objects [5]. The thresholding method is that the easiest way to segment the x-ray image. B. Median Filter Median filtering may be a nonlinear signal process technique that is beneficial for noise reduction in pictures [6]. It's the known order-statistic filter that replaces the worth of a pixel by the median of the intensity levels within the neighborhood of that pixel. C. Thresholding Threshold is one in every of the best segmentation approach. It will take out the item from the surroundings by grouping the intensity consistent with the brink price [7]. D. Edge Detection Edge detection may be a way of image segmentation techniques that find out the presence of a grip or line in a picture and express them in the level of acceptable means [8]. The function of edge recognition is to alter the image information to reduce the quantity of information to be processed [9]. E. Region of Interest A region of interest, may be a selected by set of samples inside a data set that is known for a specific purpose. The formation of unite measures in ROI is often employed in several application areas. In medical imaging, the boundaries and its limitations of cancerous cells could also be outlined on a picture or during the activity for discovery its size [10] and its belongings area component that is concerned with mathematical and geometrical options. First Order Statistics 179

3 First-order statistics live the chance of observing a gray value at a randomly-chosen location within the image. First-order statistics are computed from the bar graph of pixel intensities within the image. These rely solely on individual pixel values and not on the interaction or co-occurrence of neighboring pixel values [11]. Where, i is between (0, 255), p (i) the likelihood density of incidence of the intensity, as determined from the histogram with total range of pixels within the image is given as[12]. Where i= 0, 1, 2. G-1 G=gray level tone of a picture. N=number of cell in horizontal domain M= range of cell in vertical domain The textures that obtained from the histogram: A. Energy (1) It presents the total of square pixel elements. It's additionally referred to as uniformity or the angular moment that is range from 0 1 and varies from zero to one. Energy is E for a continuing image [13]. B. Entropy (2) It measures the randomness of a gray-level distribution. The Entropy is predicted to be high if the grey level area unit spread indiscriminately right through the image [14]. It s specified by C. Mean H= - (3) H = - total (p.*log2(p)) It supplies the mean of the grey levels inside the image. The Mean is calculated to be enlarging the level of pixels, if the total sum of the grey levels of the image is high [13]. It s given by µ = - p D. Variance (4) It s concluded that the variance is the distribution of grey levels pixels. The Variance is expected to be further large if the grey levels of the image area are expended up significantly [14]. It s given by σ2 (5) E. Standard Deviation Standard Deviation describes a large amount of variation or dispersion get from the average of it [14]. 180

4 F. Skewness (6) Histogram display skewness that is additionally referred to as third moment. It calculates the zero for symmetric histograms, positive by histograms inclined to the correct and right (about the mean) and negative for histograms inclined to again the correct and right to the left. It meant for smooth and glossy image, the results show the value involves being negative. Following equation is employed to calculate third moment [14]. µ3= σ-3 (7) G. Kurtosis The kurtosis is the histogram of uniformity [12]. Kurtosis is the level of distribution; summarized as a fourth middle moment of the normalized distribution. There are several features of kurtosis typically come across, together with the kurtosis accurate [14, 15]. Extracting Respiratory Organ (8) In the sample pictures, the background illumination is obvious within the center of the image than at the opposite regions. The morphological gap operation is employed to estimate the background illumination. Morphological in erosion followed by dilation, the process has the impact of eliminating the object, which will not utterly contain the structuring part. To form a further unvarying background, work out the conditions of that image, from the first image, once subtraction, the image incorporates a uniform background however is currently too dark. The thresholding has been applied with adding a pseudo colour image a matrix that is RGB to spot every object within the matrix image with a unique colour in the associated colour map matrix. Fig 1 shows X-ray Fig 2 shows identified Fig 3 shows histogram of of cancer patient cancer effected area cancer effected X-ray 181

5 Fig 4 shows X-ray of Fig 5 shows identified Fig 6 shows histogram of cancer patient cancer effected area cancer effected X-ray Fig 7 shows X-ray of Fig 8 shows identified Fig 9 shows histogram of cancer patient cancer effected area cancer effected X-ray Image Fig 1 Fig 4 Fig 7 Mean Variance Energy Skewness 5.67 e e e- 005 Kurtosis 6.8 e e e- 007 Entropy Standard Deviation Table 1 Statistical Features Values Fig 10 shows graphical representation of Energy v/s Entropy The statistical features for three X-ray images are exposed in the table-1 taken after three check-out of cancer patients taken by the X-rays and also highlighted the affected area of cancerous cells. The graphical representation shows in fig 10 that the respiratory organs present additional symmetrical form, where the entropy is decreased and the energy is increased. 182

6 Conclusion International Journal of Computing Academic Research (IJCAR), Volume 4, Number 6, December 2015 This search aim to extract the cancer from the respiratory organ by applying the segmentation methodology and median filter employed the region of interest and extract the cancer out from the remainder respiratory organ s image. The comparison between components related to statistical features as shown in the table-1. The skewness is high as shown in fig 7, this means that the feel of the respiratory organ become additional symmetry than before the energy increase and therefore the entropy decrease. That means the quantity of grey level pixels within the x-ray image increase energy of grey level, the tissue of the respiratory organ become lumpy after disposed the treatment. References [1] Ashnil Kumar et. al., A Graph-Based Approach to the Retrieval of Volumetric PET-CT Lung Images, International Conference of the IEEE EMBS, San Diego, California, USA, 01 September [2] Imran Fareed Nizami et. al., A Wavelet Frames + K-means based Automatic Method for Lung Area Segmentation in Multiple Slices of CT Scan, ISBN: /14/$ IEEE [3] Shi H. Lim et. al., Respiratory Motion Tracking using the Kinect Camera, IEEE Conference on Biomedical Engineering and Science, ISBN: /14/$ IEEE [4] S. Ahdi Rezaeieh et. al. Pleural effusion detection system using wideband slot-loaded loop antenna, Electronics Letters, pp , 23rd July 2015 Vol. 51 No. 15 [5] Shajy et. al., Segmentation And Feature Extraction Of Sputum For Early Detection Of Lung Cancer, First International Conference on Computational Systems and Communications, Trivandrum (ICCSC), December 2014 [6] Anjali Kulkarni1 et. al., Classification of Lung Cancer Stages on CT Scan Images Using Image Processing, IEEE International Conference on Advanced Communication Control and Computing Teclmologies (lcaccct), ISBN: /14/$3l IEEE [7] Jaspinder Kaur et. al., An automatic CAD system for early detection of lung tumor using back propagation network, 2014 International Conference on Medical Imaging, m-health and Emerging Communication Systems (MedCom), /14/$ IEEE [8] Sayali Satish Kanitkar et. al., Detection of Lung Cancer Using Marker-Controlled Watershed Transform, 2015 International Conference on Pervasive Computing (ICPC), /15/$ IEEE [9] Vinod Kumar et. al., Detection system for lung cancer based on neural network: X-Ray validation performance, International Journal of Enhanced Research in Management & Computer Applications, Vol. 2 Issue 9, Nov.-Dec., 2013, pp: (40-47), ISSN: [10] Vinod Kumar et. al., Neural Network Based Approach for Detection of Abnormal Regions of Lung Cancer in X-Ray Image, International Journal of Engineering Research & Technology (IJERT), Vol. 1 Issue 5, July 2012, ISSN: [11] Dr. S. A. Patil et. al., Lung Cancer Classification Using Image Processing, International Journal of Engineering and Innovative Technology (IJEIT) Volume 2, Issue 3, September , ISSN:

7 [12] Vinod Kumar et. al., Pulmonary Nodules Diagnosis from X-ray Imaging using Image Processing, International Conference on Recent Trends of Computer Technology in Academia (ICRTCTA) Janardan Rai Nagar Rajasthan Vidyapeeth University, Udaipur-India, April 2012 [13] D. Hadházi et. al., Lung Nodule Volume Measurement using Digital Chest Tomosynthesis, /15/$ IEEE [14] Jhilam Mukherjee et. al., Automatic Detection and Classification of Solitary Pulmonary Nodules from Lung CT Images, /14 $ IEEE, DOI /EAIT [15] Keisuke Yokota et. al., Automatic Detection of GGO Regions on CT Images in LIDC Dataset Based on Statistical Features, SCIS&ISIS 2014, Kitakyushu, Japan, December 3-6, 2014, IEEE. 181